The effect of 18F-florbetapir dose reduction on region-based classification of cortical amyloid deposition

There are specific dose recommendations for diagnostic amyloid PET imaging with 18F-florbetapir, but they may not apply to research studies using regional quantitative analysis. We, therefore, studied the effect of tracer dose reduction on the discriminative power of regional analysis.

Using bootstrap resampling of list-mode data from 18F-florbetapir scans, a total of 800 images were reconstructed for four different dosage levels: 100, 50, 20, and 10%. The effect of the injected dose on the variation of measured radiotracer uptake was determined in large cortical regions defined on co-registered and segmented magnetic resonance images. The impact of the observed variation on the discrimination between normal controls and patients with AD was then assessed using data in a cohort study described by Fleisher et al. (Arch Neurol 68(11):1404-1411, 2011).

The coefficient of variance for the cortex to cerebellum uptake ratio increased from 0.9% at full dose of 300 MBq to 2.5% at 10% of this dose, but was still small compared to biological variation. It, therefore, had very little impact on discrimination between AD and elderly controls. The original area under the ROC curve was 0.881, decreasing to 0.878 at 10% of full dose. Original sensitivity for discrimination between AD and controls was 82.0%, while specificity was 77.3%; these decreased to 81.8 and 77.1%, respectively, at the reduced dose. However, the number of subjects within the classification border zone between proven amyloid pathology and young healthy controls increased substantially by 7 to 14%.

A substantial reduction of tracer dose increases uncertainty at the classification border zone while still providing good discrimination between AD patients and controls when using activity data from cortical regions defined on co-registered and segmented MR scans.